Okay, I have basic needs here, set projects presented for a photography site. My existing site, built over fifteen years with Rapidweaver, plugins, scripts and themes, would be loaded up about once a year with the usual issues that presents, a Jenga tower of everything needing updating, clashes emerging, all that. I had, for ages, resolved to make it simpler, to redesign for consistency and ease of adding new projects. I had also resolved to have no addons, just the basics.
As mentioned elsewhere on this site, issues emerged with Rapidweaver and the wide ecosystem around the application, with factions emerging. I decided to move away and stuck with a new offering from the core team, Elements, a very clean, modern app for building sites. A year passed and I did nothing with it, the task of redesigning needed at least a week and I just never had the time. The latest version offered an MCP server, opening the option of agentic support and I gave it a go.
I pointed Claude at my existing site, told it to grab the text and images and to assemble a site with a similar aesthetic in Elements. It asked me a few questions and set to work. Within minutes I had an option in front of me, one I liked. I caught up on a novel I was reading while it worked away, yep, I read a novel, glancing over to see if there were any issues it raised. Some I ignored. Some I dealt with myself, some I asked it to propose a solution. I ran through the final finishing myself, more a chance to check things through than anything. The whole thing was done with minor attention from me, over two days. Really just me dipping in every now and then.
If this is where we are starting, it’s going to be very interesting as things progress.
I was in the very first group of beta testers but didn’t really have the time (or motivation) to learn something entirely new (and in many ways more complex). I let my subscription expire but downloaded the first AI version and set up Claude via the MCP server.
Using Indesign I made a quick layout of something I liked, took a screenshot, then fed it to Claude and asked it to make a site (it was just a single page). Whilst I haven’t had time to follow it up, it did a fairly decent job of reproducing it, and the next test will be to point it to my existing site (Rapidweaver Classic) and see if it can reproduce it.
Website design is not something I particularly enjoy, but I can see the new AI options making it far more enjoyable.
I recently found myself in a similar position. My camera club has a moderately sophisticated site, which as well as the usual announcements handles our monthly competitions (in which as well as the judge’s scores, we allow members to vote), special interest groups and so on. The usual off-the-shelf club sites wouldn’t come anywhere near what we need. The site was put together a while ago by a professional designer, a member of the club (and a fine photographer). It was based on old, outdated frameworks. He had intended to update it but sadly died, very young, last year. We were faced with a likely bill, judging by various designers we consulted, of at least £5,000 for analysis, update and maintenance; and we simply don’t have that kind of money.
I spent a few days feeding ChatGPT all the files it asked for (and helped me find, buried in a very complex file structure). It then produced a 29-page specification, which I couldn’t fault. We (funny how one finds oneself teaming up with AI and using plural pronouns, isn’t it?) have now produced an updated site, written from scratch, using up-to-date frameworks. It’s not yet fully finished, but it does everything the old site did and more. The total cost, apart from my freely-given time, has thus far been £20. I expect a second month’s subscription to ChatGPT will be necessary before it’s ready to put into use.
I’ve been writing software, on and off, for 50 years and I’m hard to impress but I have found this experience quite staggering. It’s saved the club from what would have been a very difficult call on its members. But if I were a newly-qualified web designer, I’d be seriously worried.
I’ve been thinking about the possibility of using A.I. to build a simple group membership tracking website using a traditional LAMP (Linux-Apache-MySQL-PHP) stack or something similar.
Many years ago, our group had a self-built web-enabled FileMaker site that covered our needs nicely, but eventually it became too expensive/complex to maintain. There also were third party tools that handled the job well at low cost, but they eventually disappeared, often due to misguided venture capital ambitions.
Our main issue with running a LAMP site was on the design side, not the backend side. Ok, a little bit on the maintenance side, too. We don’t need anything fancy, but it should look reasonably attractive and follow common UI conventions.
I suspect a lot of pricey membership management website vendors are about to experience Clayton Christensen’s “low-end disruption” phenomenon soon:
Disruptive Innovations are NOT breakthrough technologies that make good products better; rather, they are innovations that make products and services more accessible and affordable, thereby making them available to a larger population.
I was in the same situation a few years ago with RapidWeaver vs. Elements vs. all the others. I decided I did not want to invest time learning something new so I ended up working with one of the RW plugin developers to replace an obsolete plugin.
I will not be updating my site again. When it breaks, I will pull it down. I would rather spend the time shooting photos than managing a photo web site – but your experience with Claude makes me wonder if that might be a good choice in the future.
Although I didn’t use Claude (I used MS Copilot) my first attempt at “vibe coding” was equally successful and gratifying.
I’ve been using GeburtstagsChecker for at least a decade. It’s my go-to for remembering birthdays, anniversaries, and other milestones. The Mac version has the ability to load from an external file (mine contains maybe 250 entries), while the iOS version only reads birthdays from the Contacts app.
I’ve long wanted a way to display the more extensive (and flexible) Mac text data from my iPhone. But that feature doesn’t fit Earthlingsoft’s vision. It took me all of 15 minutest to have Copilot create a browser-based, statically-served javacript webapp that reads an uploaded data file and presents the same info that the Mac app does.
I’ve got a lot of programming experience, but I’ve never learned Javascript, and I don’t have any desire to do so. I’m delighted.
That was then. I’ve heard that the relentless hammering of sites that the LLM scrapers engage in has wreaked havoc on a lot of self-hosted sites. Note that the LLM vendors often ignore robots.txt and actively try to circumvent other access controls. They can bring your site to its knees, or drive your hosting costs through the roof.
That is correct. It has impacted several organizations that I am involved with. It does take some planning and effort to mitigate what can end up being a denial-of-service attack. It also takes very careful consideration of one’s security model if private and/or personally identifiable information is made available behind an authentication wall. I may have mentioned here that my alma mater removed its directory from view as a precaution after many years of public availability.